Recently, the highly anticipated "2025 International Artificial Intelligence Security Report" was officially released, representing the collective wisdom and dedication of 96 top AI experts from around the world. The report provides a comprehensive assessment of the potential capabilities, risks, and risk management methods of Artificial General Intelligence (AGI), offering valuable insights for the safe development of AI globally.
In the report, experts conduct a thorough and in-depth evaluation of AGI's capabilities. They believe that with continuous technological advancements, AGI will play an increasingly important role in the future, with its intelligence level rising and its application fields broadening. However, alongside these benefits, AGI also poses a series of potential risks. These risks include, but are not limited to, malicious use, system failures, and systemic risks, which may have far-reaching impacts on individuals, society, and even the globe.
Regarding malicious use, AGI could be exploited by criminals to harm individuals or destabilize society through the generation and dissemination of false content. For example, AI technology could be used to create fake news, rumors, or malicious information to mislead the public or create social panic. Additionally, the reliability of AGI cannot be overlooked. Due to the complexity and uncertainty of the technology, AGI may experience various failures or errors during use, negatively impacting the sectors it serves.
Beyond malicious use and reliability issues, AGI may also trigger systemic risks. Due to the interconnectedness and dependence of AI systems, a failure in one system could trigger a chain reaction in others, potentially leading to the collapse of the entire system. This risk is particularly severe in critical sectors such as finance, transportation, and healthcare, where problems could result in significant societal losses.
To address these risks, the report also proposes corresponding risk management methods. However, experts point out that current risk management methods face numerous challenges both technically and socially. Technically, existing risk management methods may not fully cover all potential risk points, especially when dealing with complex and ever-changing AI systems. Socially, the implementation of risk management methods requires support and cooperation from all parties, but in reality, due to conflicts of interest, differing perspectives, and other reasons, effective cooperation and coordination are often difficult to achieve.
Therefore, experts call for enhanced cooperation among countries and regions worldwide to jointly address the challenges posed by AGI. They recommend strengthening technological research and development to improve the security and reliability of AI systems; enhancing legal and regulatory frameworks to clarify the norms and responsible entities for AI technology use; and improving public education and awareness to increase societal understanding and acceptance of AI technology.
